The short version: most small businesses using ChatGPT or Claude for their marketing have accidentally adopted the same voice, the same three sentence structures, and the same slightly cheerful, slightly hollow tone as every other business doing the same thing. It's not the tool's fault. It's that almost nobody bothers to feed it enough of their own real writing before asking it to sound like them. Fixing it takes about two hours, not a rebrand.
The email that gave it away
Last spring I was helping a client, an accountant in Kent I'll call Sarah, tidy up her client onboarding emails. She'd used AI to write a "welcome to the practice" sequence and was proud of it. It was warm, it had a friendly em-dash-free rhythm, it hit "we're here to support you every step of the way" and "let's make this a smooth journey together."
The problem: I'd seen that exact tone, almost word for word, from a marketing consultant client the week before, a dog groomer three weeks before that, and a web design agency around Christmas. Different industries, same fingerprint. Not the same sentences, but the same shape. Same three-comma sentences. Same closing line about being "excited to work together." It read like it had been written by someone who'd never met any of these businesses and honestly, it had.
That's the bit nobody tells you when they sell you on AI for marketing: the default voice of ChatGPT, out of the box, is a specific, learnable, recognisable style. Once you've read a hundred pieces of AI-written business content, you start spotting it in about four seconds. So do your customers, even if they can't name what's wrong.
Why this happens (it's not random)
Large language models are trained to produce the statistically most likely, broadly acceptable version of a sentence. Ask for "a friendly welcome email for new clients" with no other input and it will reach for the most common patterns in its training data, which skews towards corporate blog copy, marketing templates and SaaS onboarding flows. That's the average of everything, which is precisely why it sounds like nothing in particular.
This is a textbook default bias problem, the kind Richard Thaler spent his career studying. People overwhelmingly stick with whatever option requires the least effort, even when a slightly better option is sitting right there. With AI writing, the "default" is accepting the first draft. Most business owners do exactly that because rewriting feels like extra work on top of work they were trying to save time on in the first place.
I'll say the uncomfortable part plainly: telling people to "just add personality to the prompt" doesn't solve this, and most of the advice floating around does exactly that and stops there. Typing "write this in a bold, cheeky tone" gets you a caricature of boldness, not your voice. The model still has no idea how you write, so it substitutes a stereotype of the tone you asked for. You end up sounding like a Halifax car insurance advert. That's not a fix, it's a different flavour of generic.
The two-hour fix that changes the output
The businesses whose AI content doesn't sound like everyone else's have done one thing differently: they've fed the model a proper sample of their own real writing before asking for anything new. Here's the exact process I now run with consulting clients, and it takes under two hours the first time, then about fifteen minutes to maintain.
- Step 1: Pull 8 to 10 pieces of writing that are yours, not AI-assisted. Old emails to clients, a LinkedIn post you wrote before any of this existed, a voicemail transcript, a text you sent a supplier when you were annoyed. Aim for at least 1,500 words total.
- Step 2: Paste all of it into one document and ask the AI to analyse it for sentence length, favourite words, where you use humour, where you're blunt versus where you soften things, and what you never say. Get it to write this up as a "voice profile" document, roughly 300 to 500 words.
- Step 3: Save that profile and paste it into every new content prompt, every single time, with an instruction along the lines of "match this voice exactly, including sentence length and any bluntness."
- Step 4: Read the first output aloud. If you wouldn't say it out loud to a client sat opposite you, cut it, even if it sounds "professional." Professional is often the disguise generic wears.
- Step 5: Build a short "banned words" list based on what keeps recurring that isn't you: things like "," "elevate," "unlock," "game-changer." Add to it weekly for the first month.
Sarah did this with her welcome sequence. The rewrite kept two sentences from the original draft and replaced the rest with things closer to how she talks to clients on the phone, including one line that says "I'll chase HMRC so you don't have to, and I'll tell you honestly if something's going to be a headache." Open rates on that email sequence went from 41% to 58% over the following quarter. Not because the offer changed. Because it stopped reading like a template.
What distinct voice buys you
This isn't just an aesthetic complaint about AI slop. Voice is one of the last defensible bits of a small business, because it can't be copied by a competitor who undercuts you on price. Look at the brands that have got this right for decades rather than months: Disney's voice is unmistakable in a single sentence, Headspace built an entire category around a calm, slightly wry tone that never once sounds like a wellness brand pretending to be calm, and GoDaddy went from cringeworthy to warm once they stopped chasing shock value and found an actual consistent character. None of that happened by accident, and none of it happened by asking a tool for "a friendly tone." It happened because someone decided, in detail, what the brand does and doesn't sound like, the same discipline Coco Chanel applied to everything with her name on it, where refusing to follow trends became the entire signature.
You don't need Disney's budget to get this right. You need the discipline to write down what you sound like before you let a machine write for you.
Where this shows up most (and costs you the most)
Three places I see the sameness problem doing real commercial damage:
- LinkedIn posts. The "I used to think X, then I learned Y" format is now so overused it's become a running joke, and prospects scroll past it without reading a word.
- Website copy. Homepages that open with "In today's fast-paced world" or "we understand that running a business is hard" tell a visitor nothing and read as if written by nobody in particular.
- Sales emails. If your follow-up email could be sent by literally any consultant in your niche without changing a word, it will get the same reply rate as literally any consultant in your niche: low.
A useful sense check: instead of another block of AI text, try something interactive that only your business could have built, the way an interactive calculator forces a visitor to engage with your specific numbers rather than skim your specific adjectives. It's much harder to sound generic when the content is doing actual work for the visitor rather than just describing how helpful you are.
The bit most people skip
Here's what I won't pretend isn't true: building a real voice profile is more effort than typing "make this sound friendlier," and most business owners, even ones who know all of the above, will keep taking the easy default because they're busy and the AI output is "good enough." That's a legitimate choice if you're a solo operator sending three emails a week. It stops being a legitimate choice the moment you're relying on content to win business against competitors who've done the two hours of work. Sameness is invisible until you're the one being scrolled past.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
If you'd rather have someone build the voice profile and the prompt system with you rather than trial and error it alone, that's exactly the kind of hands-on work I do with clients through AI implementation coaching, usually a two to three hour session that leaves you with a document you'll use for years, not a one-off blog post.
A quick test you can run this week
Take a piece of AI-written content you've published in the last month. Read it next to a competitor's AI-written content in the same niche. If you could swap the company names and nobody would notice, you've got the sameness problem. If a regular reader could tell which one was yours without seeing the logo, you're already ahead of most of the small businesses currently publishing on autopilot.
Frequently asked questions
Does AI-written content sound worse than human-written content?
Not worse exactly, just averaged. AI models default to the statistically most common phrasing across their training data, which tends to be safe, mid-tone corporate copy. It reads fine but forgettable, and forgettable is the actual cost, not quality.
How long does it take to fix a generic AI voice?
Building a proper voice profile from your own past writing takes around two hours the first time, then roughly fifteen minutes to update each month as you add new examples and phrases you want banned from future output.
Can I just tell ChatGPT to "write with personality" instead?
You can, but it usually produces a stereotype of the tone you asked for rather than your actual voice, because the model has no real sample of how you write. Feeding it your own past emails and posts works far better than describing a tone in the abstract.
Is this only a problem for solo business owners, or does it affect bigger companies too?
It affects anyone using AI for marketing content without a defined voice document, regardless of size. The difference is that bigger competitors with dedicated marketing teams tend to catch it and fix it faster, which is exactly why small businesses can't afford to skip the step.
Want this done for you? See AI automation for coaches and consultants.
Related reading: Your Business Now Has an AI Reputation, and You Probably Haven't Checked It and Starbucks Marketing Strategy: How They Built a Brand That Wins.
This overlaps with my main post on the topic: ai marketing sameness small business.
